Differential Effects of Heterochromatin Protein 1 Isoforms on Mitotic Chromosome Distribution and Growth in Dictyostelium discoideum

Plasmids.

To generate C-terminally green fluorescent protein (GFP)-tagged HcpA and HcpB, the respective cDNA was amplified using primers 5′ GAATTCAAAATGGGAAAAAGAGATAAGAG and 5′ GGATCCACTTTGTTGACCCTTATAACC for hcpA and 5′ GCTAAAAGAATTCAAAATGGGAAAAAGAG and 5′ GGATCCACTTGGCTGACCACTATAACC for hcpB and cloned into the EcoRI and BamHI sites of pDd-GFP (36). For N-terminal GFP fusions, hcpA was amplified using primers 5′ GTCGACATGGGAAAAAGAGATAAGAGAATAATAG and 5′ GGATCCTTTTAACTTTGTTGACCCTTATAACC, and hcpB was amplified with primers 5′ CAGCTAAAATTGTCGACATGGGAAAAAGAG and 5′ GGATCCTTAACTTGGCTGACCACTATAACC and cloned into the SalI and BamHI sites of pdneo2-GFP (56). For the HcpA-red fluorescent protein (RFP) fusion, the GFP coding sequence of pDd-HcpA-GFP was replaced by mutant RFP (RedStar) from p415Gal1-Redstar (30) by digestion with BglII/XhoI and ligation into BamHI/XhoI-digested pDd-HcpA-GFP.

The chromoshadow domain deletions (HcpAΔC-GFP and HcpBΔC-GFP) were designed with primers 5′ GAATTCAAAATGGGAAAAAGAGATAAGAG and 5′ ATTGGATCCACCTTCAATATAAGGTACACC for hcpA and 5′ GCTAAAAGAATTCAAAATGGGAAAAAGAG and 5′ ATTGGATCCACCTTCAATATAAGGTACACC for hcpB and cloned into the EcoRI and BamHI sites of pDd-GFP. For these constructs, identical reverse primers for both isoforms were used, since the primer binding sites were identical in both coding sequences.

Knockout constructs were generated using a blasticidin resistance cassette (68) with flanking arms of 625 bp and 501 bp for hcpA or 703 bp and 821 bp for hcpB. Primers to create recombinogenic arms were 5′ ATAGGATCCGAAGACCCTTTTAAGAAATGTTGTC and 5′ TTGAAGCTTTTTTCTTGTTCTCTTGAACTTTC as well as 5′ GCATGCCAAATCATTTAATTGATGATG and 5′ ATGGGCCCGAAGACGAATTATTTAGTAATTCTTTAAAATATG for hcpA, and 5′ ATTGGGCCCGAAGACAAAATTCACCATATAAGGGG and 5′ ATTGCATGCCCATGTATCTTAATCTGATG as well as 5′ ATTAAGCTTCATCAACAACAGCAGCAGCACC and 5′ ATTGGATCCGAAGACTTACCCCATCCAAACAATGAG for hcpB.

For both targeting constructs, the recombinogenic arms were cloned into pGEM7z-BS^R, flanking the blasticidin resistance cassette. The vector pGEM7z-BS^R was generated by cloning the BS^R cassette (68) into the HindIII/XbaI sites of pGEM7z (Promega).

Knockout constructs were designed to disrupt the respective open reading frames (ORFs) between the N-terminal chromodomain and the nuclear localization signal (NLS) within the hinge region (data not shown). The truncated genes should thus not generate any functional, correctly localized protein.

For disruption of both endogenous hcp genes, the BS^R cassette of both targeting constructs was excised with XbaI and HindIII and replaced by a BS^R cassette with flanking loxP sites excised from pLPBLP (9) with BcuI and HindIII.

Recombinant His-tagged HcpA and HcpB was produced by cloning the cDNAs into the NdeI and BamHI sites of pET15b (Invitrogen). Primers used for amplification of the respective cDNAs were 5′ CATATGGGAAAAAGAGATAAGAGAATAATAG and 5′ GGATCCTTTTAACTTTGTTGACCCTTATAACC for hcpA and 5′ GCTAAAATTGTAACATATGGGAAAAAGAG and 5′ GGATCCTTAACTTGGCTGACCACTATAACC for hcpB.

Induction and protein expression was performed in Escherichia coli BL21(DE3) according to the manufacturer's instructions.

For expression of 6× His-tagged fusion proteins in Dictyostelium discoideum, the coding sequences were amplified from the respective pET-15b-derived plasmids and cloned into the EcoRI and BamHI sites of pDEX-DnmA, from which the DnmA (32) coding sequence had been excised.

Strains.

Transformations to generate knockout and overexpression mutants were performed in the Ax2 parent strain. Dictyostelium transformations and cotransformations were carried out as described previously (47, 48).

Gene disruptions and subcloning were done by homologous recombination (74). Clones were analyzed by PCR as described previously (38).

For hcpA, primers 5′ CAGTTACTTGTTTCATTATGGC and 5′ CGCTACTTCTACTAATTCTAGA were used to verify site-specific integration of the BS^R cassette. For hcpB, primers were 5′ AAAAGATATAGAATCTACAACTATC and 5′ CGCTACTTCTACTAATTCTAGA.

For the generation of hcpA/hcpB double knockout strains, the Ax2 strain was first transformed with the hcpA targeting construct containing a BS^R cassette with flanking loxP sites. Isolation of hcpA mutant cells was carried out as described above. Transient expression of Cre-recombinase, screening for blasticidin, and G418 sensitivity were carried out as described previously (9). The generated hcpA^lp strain was then transformed with the hcpB targeting construct or transformed with pDEX-RH-His-HcpA prior to hcpB targeting.

RT-PCR analysis.

Total cellular RNA was prepared as described previously (37). Total RNA was DNase-treated before cDNA synthesis to eliminate genomic DNA contaminations. Reverse transcription-PCR (RT-PCR) analysis using Revert Aid H Minus M-MulV Reverse Transcriptase (MBI) was performed according to the manufacturer's instructions. Primers used to analyze expression levels of Hcp isoforms and the unrelated thioredoxin gene in knockout and overexpression strains were hcpA forward (for), 5′CATATGGGAAAAAGAGATAAGAGAATAATAG; hcpA reverse (rev), 5′ GGATCCTTTTAACTTTGTTGACCCTTATAACC; hcpB for, 5′ GCTAAAAGAATTCAAAATGGGAAAAAGAG; hcpB rev, 5′ GGATCCACTTGGCTGACCACTATAACC; trx for, 5′ GAACGAGCTCCATGGCCAATAGAGTAATTCATG; and trx rev, 5′ CGCGGATCCTTATTTGTTTGCTTCTAGAGTACTTC.

Cell culture.

Growth curves of Ax2, hcpA mutant, and hcpB mutant were done in axenic suspension culture in HL5 medium (67) on a rotary shaker at 150 rpm. For the overexpression strains, HL5 medium supplemented with 20 μg/ml geniticin was used. The starting cell density was 1 × 10⁵ to 3 × 10⁵/ml for growth at 22°C and at 15°C. Cell densities were measured with a Coulter Counter every 4 to 6 h for 48 to 72 h at 22°C and every 24 h for 10 to 15 days at 15°C. All growth curves were measured in duplicate.

In order to determine cell viability, the plating efficiency of cell suspensions that had been cultured for 48 h on rotary shakers was determined. Cell densities were measured before serial dilution in phosphate buffer. Cells were plated together with Klebsiella aerogenes suspension on nonselective SM agar (67) plates and allowed to grow for 72 h. Colonies were counted, and the plating efficiency was calculated by dividing colony number by cell density times dilution factor.

Immunodetection.

Dictyostelium discoideum cells were grown on coverslips in petri dishes containing the appropriate selective medium for 20 to 24 h. Cells were then fixed in −20°C methanol for 20 min, washed three times in 1× phosphate-buffered saline (PBS), and blocked with PBG buffer (1× PBS with 3% bovine serum albumin and 0.045% cold water fish gelatin; Sigma) for 1 h at 37°C. Primary antibodies were applied in appropriate dilutions and incubated over night at 4°C.

For detection of histone H3K9 dimethylation, a polyclonal rabbit antibody (Upstate Biotechnology) diluted 1:100 was used. For staining of centrosomes, the anti-DdCP224 antibody (18) was used in a 1:100 dilution. After six washes in PBG, Cy3-conjugated secondary antibodies (Dianova, Hamburg, Germany) were applied in a 1:1,000 dilution in PBG for 1 h at 37°C. DNA was stained with 4′,6′-diamidino-2-phenylindole (DAPI) (1 mg/ml) diluted 1:15,000 in 1× PBS.

Mitotic analysis.

For mitotic assays, cells were precultured in petri dishes at 22°C and 15°C and then diluted to 3 × 10⁵/ml at 22°C and 7 × 10⁵/ml at 15°C, respectively. Cells were then grown on coverslips in petri dishes covered with 5 ml of HL5 medium for 20 to 24 h and fixed as described above. Mitotic cells were detected by anti-DdCP224 or by anti-α-tubulin staining. For determination of the frequency of anaphase bridges, only late mitotic cells with spindles that were ≥5 μm were counted, and the number of mitoses with DNA bridges was determined.

Statistical evaluation.

To get a significance level for the observation that overexpression of GFP-HcpA leads to a higher frequency of anaphase bridges in late mitotic cells, two procedures were applied.

Procedure 1.

As the number of anaphase bridges also depended on the number of counted cells, we adjusted the data to 90 counted cells per cell type, resulting in 11 or 12 anaphase bridges for GFP-HcpA cells and letting all other numbers be fixed. This procedure even reduced the chance to obtain a convincing level of significance. We computed the chi-squared (χ²) statistics with the following equation: [k_i − (m/8)]²/(m/8), where the k_i is the respective number of anaphase bridges, with k₅ = 12 or 11 and m = 29 or 28 for the total count of anaphase bridges. The calculated value was compared with the respective values obtained after randomly distributing m objects in 8 boxes, the well-known urn model in stochastics, which represents the null hypothesis we want to reject. Thus, for k₅ = 12, we obtained the significance value of α = 0.0014; for k₅ = 11, the significance value was α = 0.0052.

Procedure 2.

In each of the first seven cases, we have purely randomly chosen 90 trials and noted the k_i numbers of the chosen trials in each of the seven cases in which there occurred anaphase bridges. Of course, k₈ = 3. Like the first procedure, we determined the significance level (α) of the obtained 8-tuple (k₁ to k₈). This was repeated 1,000 times. For the mean of the obtained values for the significance level, we obtained an α_m of 0.0032 and for the median an m of 0.0009 (see Table 3).

Summarizing these numerical results, our experimental observations show for the significance level of at least 0.005 that overexpression of GFP-HcpA results in a higher frequency of anaphase bridges than in the other cell types.

Fluorescence microscopy.

For microscopic analysis a Leica DM IRB inverted fluorescence microscope was used, and for image acquisition a Leica DC 350F digital camera and IM50 software were used. Images were processed in Adobe Photoshop.

Electron microscopy.

Nuclei were isolated from HcpA-GFP- and HcpB-GFP-expressing cells (33) that had been treated with 100 μM thiabendazole for 4 h. The nuclei were centrifuged onto round coverslips and fixed with 2% paraformaldehyde in modified PHEM buffer [15 mM piperazine-N,N′-bis(2-ethanesulfonic acid), 6 mM HEPES, 10 EGTA, 0.5 mM MgCl₂, pH 7] (24) for 20 min. The preparations were treated with 1 mg/ml bovine serum albumin, 5% fetal calf serum, and 2% cold water fish gelatin to reduce nonspecific staining and were incubated with an anti-GFP antibody for 1 h. Five nanometer protein A gold (Department of Cell Biology, University Medical Center, Utrecht, The Netherlands) was applied for 2 h, and after extensive washes the nuclei were postfixed with 1% glutaraldehyde for 5 min. The subsequent fixation with 1% OsO₄ in cacodylate buffer, dehydration, and embedment in Epon/Araldite were carried out according to standard procedures (27). Thin sections were cut using a Reichert Ultracut E, stained with uranyl acetate and lead citrate, and observed in a Jeol EM 1200 EX II.

Pull-down assays.

His-tagged HcpA and HcpB were purified on Ni-Sepharose beads (Amersham) according to the manufacturer's instructions.

Cell lysates from a Dictyostelium discoideum strain expressing GFP-tagged fusion proteins were prepared by sonifying 1 × 10⁸ cells from axenic suspension culture three times for 15 s in 10 ml lysis buffer (20 mM phosphate buffer, pH 7.5, 300 mM NaCl, 10 mM imidazole, 0.5% NP-40).

For isolation of Hcp binding proteins, recombinant His-tagged HcpA or HcpB was immobilized on a Ni-Sepharose column. Cell lysates were mixed with precoated Ni-Sepharose beads and rotated for 1 to 2 h on a spinning wheel. Beads were spun down by brief centrifugation. The supernatant was stored, and the beads were washed with at least 20 column volumes of binding buffer. The respective Hcp and bound proteins were then eluted with 250 mM imidazole. Eluted fractions were analyzed by sodium dodecyl sulfate-polyacrylamide (12%) gel electrophoresis. Proteins were stained with Coomassie and detected on Western blots by a monoclonal anti-GFP (Chemicon, Temecula, CA) or anti-His antibody.

The same protocol was applied for determining in vivo interactions, except that extracts from His-HcpA/GFP-HcpB- and His-HcpA/GFP-HcpA-cotransformed cell lines were directly incubated with the Ni-Sepharose beads.

Strains.

Strains used in this study were Dictyostelium discoideum Ax2, Ax2 HcpA⁻, Ax2 HcpB⁻, Ax2 GFP-HcpA, Ax2 GFP-HcpB, Ax2 His-HcpA, Ax2 His-HcpB, Ax2 HcpA-GFP, Ax2 HcpB-GFP, Ax2 HcpAΔC-GFP, and Ax2 HcpBΔ C-GFP and E. coli DH5α (Promega) and BL21(DE3) pLysS (Invitrogen).

Nucleotide sequence accession numbers.

Sequences determined in the course of this work were submitted to Dicty Base (http://dictybase.org/) under the following accession numbers: hcpA, DDB0220646; hcpB, DDB0220645; and hcpC, DDB0220647.

HP1-like proteins in Dictyostelium discoideum.

A database search using the human HP1α sequence and a tBLASTN algorithm revealed the presence of three putative genes encoding HP1-like proteins in the Dictyostelium discoideum genome. All three displayed the characteristic hallmarks of HP1 proteins: an N-terminal chromodomain and a C-terminal chromoshadow domain separated by a less conserved hinge region. The genes were denominated hcpA, hcpB, and hcpC (Fig. 1). HcpA displays 70% identity (77% similarity) with HcpB but only 62% identity (75% similarity) with HcpC. Sequence analysis further showed that amino acid residues within the chromodomain that are required for function, such as methyl-lysine binding, were highly conserved in the Dictyostelium proteins, whereas functionally important amino acid residues within the chromoshadow domain were less conserved (Fig. 1). The HcpA chromodomain displayed highest similarity to that of human HP1γ (79%) and slightly less to the α- and β-isoforms (73% and 77%, respectively). For comparison, the chromoshadow domain of HcpA displayed 52%, 48%, and 46% similarity to the human α-, β-, and γ-isoforms, respectively.

RT-PCR analysis showed that hcpA and hcpB were expressed in axenically grown cells (see below) and throughout development (data not shown). hcpC transcripts were not detected under these conditions, while the gene-specific primers readily showed a PCR product when using genomic DNA (data not shown). To see if hcpC was induced as a potential compensatory mechanism, we investigated its transcription in hcpA and hcpB knockout strains but could not detect any RT-PCR signal. Apparently, hcpC is not expressed at all or has a very restricted expression pattern. It was therefore excluded from further studies.

HcpA and HcpB display largely identical subnuclear localization.

To study the subnuclear organization of Dictyostelium heterochromatin, we used C-terminally GFP-tagged HcpA and HcpB (HcpA-GFP and HcpB-GFP). Both proteins localized to one major spot at the nuclear periphery and several minor foci (Fig. 2A). Depending on expression levels, some nucleoplasmic staining, probably representing unbound HP1 molecules, was observed. Furthermore, coexpression of HcpA-RFP and HcpB-GFP showed that HcpA and HcpB colocalize (Fig. 2B), indicating that the subnuclear distribution of both isoforms is very similar. Electron microscopy revealed that the fusion proteins colocalize with electron-dense structures at the nuclear periphery close to the centrosome (shown for HcpA-GFP in Fig. 2C). These structures contain the kinetochores/centromeres of Dictyostelium chromosomes (43). Centrosomes have previously been characterized by electron microscopy and immunostaining with specific monoclonal antibodies, like DdCP224 (see below) (19).

Dimethylated lysine 9 of histone 3 (H3K9me2) is a further hallmark for heterochromatin and serves as a binding site for HP1 proteins (1, 34). To test for this association in Dictyostelium, double-labeling experiments were performed and showed that both HcpA-GFP and HcpB-GFP colocalized with H3K9me2 (Fig. 2A). The same was true for the respective N-terminal GFP fusion proteins (termed GFP-HcpA and GFP-HcpB; data not shown). The specificity of the H3K9me2 antibody, originally raised against methylated human histone H3, was verified by Western blotting. It recognized a single 17-kDa protein which corresponds in size to the Dictyostelium histone H3 (data not shown).

HcpA and HcpB form homo- and heterodimers in vitro and in vivo.

A conserved feature of HP1 proteins is their ability to form multimeric complexes by homo- and heterodimerization through their chromoshadow domains (3, 51). Since we found largely identical subnuclear distribution of both Hcp isoforms, we determined if HcpA and HcpB are capable of dimerization in vitro. Bacterially expressed His-tagged HcpA or HcpB was immobilized on Ni-Sepharose columns. Extracts from Dictyostelium discoideum cells expressing GFP-HcpA were passed over the matrix, and bound proteins were eluted with imidazole (Fig. 3A). Both His-HcpA and His-HcpB were able to retain GFP-HcpA and GFP-HcpB on the column (Fig. 3B). This showed that the Dictyostelium proteins could form both homo- and heterodimers. Relatively, His-HcpA retained much less GFP-HcpA than GFP-HcpB (Fig. 3B). This was reproducible in multiple experiments, which always showed a faint but defined band. The significance of the precipitation was confirmed by control experiments where beads without coupled Hcp protein were incubated with extracts from Dictyostelium discoideum cells expressing GFP-HcpA or GFP-HcpB. In both cases no band was detected in the precipitate (Fig. 3B).

GFP fusions of HcpA or HcpB lacking the chromoshadow domain (HcpAΔC-GFP and HcpBΔC-GFP) could not form any detectable dimers with immobilized His-HcpA (Fig. 3C), indicating that the chromoshadow domain is required for dimerization. This is consistent with findings for HP1 homologues from other organisms.

To confirm that both isoforms are also capable of homo- and heterodimerization in vivo, we generated cell lines that stably expressed both His-tagged HcpA and either GFP-HcpA or GFP-HcpB. Ni-Sepharose beads were then incubated with cell extracts from these cell lines (Fig. 4A). For both cotransformations, GFP-HcpA and GFP-HcpB could be coeluted with His-HcpA, indicating that both isoforms form oligomeric complexes in vivo (Fig. 4B).

Remarkably, significantly more GFP-HcpB than GFP-HcpA was coeluted with His-HcpA, although the expression levels of His-HcpA were very similar in both cell lines (Fig. 4C) and the respective GFP-fusion proteins were present in excess over His-HcpA.

In both experiments, HcpA-homodimer formation appeared to be less favorable than HcpA-HcpB-heterodimer or HcpB-homodimer formation. This suggests that the two isoforms differently contribute to heterochromatin formation.

Although the amino acid sequences of the chromoshadow domains of Dictyostelium and mammalian HP1 proteins are significantly divergent (Fig. 1), they display the characteristic hallmarks of homo- and heterodimerization. This sequence divergence possibly confers a distinct species-specific recognition of interaction partners. For example, His-HcpA formed homo- and heterodimers with the Dictyostelium isoforms but was unable to interact with murine HP1α (data not shown).

Mitotic dynamics of Dictyostelium heterochromatin.

In interphase cells, the major heterochromatic cluster is closely associated with the centrosome, a nucleus-associated body in Dictyostelium (Fig. 5, interphase; see also Fig. 2C). Since this association was stable in randomly migrating or chemotaxing cells (data not shown), this indicated a functional relationship and was reminiscent of centromeric heterochromatin in S. pombe, which is closely connected to the nuclear membrane-embedded spindle pole body (15).

To examine if the major heterochromatic cluster represented Dictyostelium centromeres, we analyzed HcpA-GFP and HcpB-GFP distribution during mitotic stages in asynchronously growing Dictyostelium discoideum cells. Mitoses were identified by antibodies directed either against DdCP224, a Dictyostelium XMAP215 homologue that stains the centrosome and the mitotic spindle (18, 19), or against α-tubulin (28).

During interphase, both Hcp-GFP fusion proteins localized to one prominent cluster at the nuclear periphery directly opposite the centrosome and colocalized with H3K9 dimethylation. In prophase, when Dictyostelium centrosomes are duplicated, the cluster split up into several smaller foci. In some preparations, up to six spots could be distinguished that may represent the centromeres of the six Dictyostelium chromosomes (Fig. 5A, b, prophase inset). This was similar to pericentromeric heterochromatin in S. pombe that also splits up during mitotic prophase (15, 55). In prometaphase, when the duplicated spindle poles separate and penetrate the nuclear envelope, the majority of the nuclear HcpA-GFP and HcpB-GFP enters the cytoplasm, indicated by a significant increase in cytoplasmic and loss of nucleoplasmic GFP staining (Fig. 5B, c). This probably represents passive diffusion and not an active process, since a similar relocalization has been described for NLS-tagged GFP (75). Both observations argue that mitosis may not be completely closed in Dictyostelium discoideum.

In metaphase, heterochromatin localizes between a bilayered structure at the central spindle that is stained by the anti-DdCP224 antibody (Fig. 5A, d) and that probably represents the kinetochore region (58). During anaphase, the heterochromatic clusters divided and moved toward the spindle poles, leading the separated DNA masses. Similar mitotic dynamics of heterochromatin were observed for histone H3K9 dimethylation (Fig. 6). Although GFP-fused Hcp proteins stay associated with heterochromatin throughout the entire cell cycle, we reproducibly detected a significant loss of Hcp proteins at heterochromatin during late mitotic stages and an increase in cytoplasmic staining compared to interphase (Fig. 5 and 6). This is best demonstrated in Fig. 5B, g, where an interphase cell with strong HcpB-GFP centromere staining is shown next to a telophase cell with very weak GFP label. Although we cannot say if this behavior represents that of the endogenous proteins, it may indicate that heterochromatic localization of Hcp proteins is dynamically regulated during the cell cycle. Similar observations for HP1 dissociation from heterochromatin during mitosis have been made in mammalian cells (13, 23). In conclusion, the colocalization data and the mitotic behavior strongly suggested that the foci stained by the GFP-labeled HP1 isoforms represent centromeric heterochromatin.

Knockout mutants of hcpA, but not of hcpB, show temperature-dependent growth defects.

To analyze the function of the two HP1 isoforms, we generated knockout strains for both hcpA and hcpB by homologous recombination (data not shown). Under standard laboratory conditions, none of the single-knockout strains displayed a significant phenotype in axenic suspension culture or during development (data not shown). However, at lower temperatures, the hcpA mutant strain but not the hcpB mutant strain displayed a slight growth defect in suspension culture compared to the parent Ax2 strain (Table 1).

In S. pombe, knockout of swi6 causes a similar temperature-dependent growth defect that is characterized by increased chromosome missegregation rates and can be mimicked by microtubule-destabilizing drugs such as thiabendazole (8). In contrast to that, we could not detect any defects in mitotic progression or an obvious chromosome missegregation phenotype. Furthermore, the temperature sensitivity of the hcpA mutant could not be mimicked by thiabendazole treatment, indicating that microtubule (dys)function was not involved (data not shown). Therefore, the reason for the temperature sensitivity of the hcpA mutant remains unclear.

Despite subtle differences at low temperature, the lack of a mutant phenotype under standard conditions argued for a redundancy of the two Hcp isoforms in most functions.

We therefore attempted to disrupt both endogenous genes to create double-knockout mutants. Since the gene disruption frequencies for the two hcp genes were reproducibly dissimilar (4% for hcpA compared to 67% for hcpB), we first created an hcpA mutant cell line that was followed by Cre-mediated excision of the blasticidin resistance cassette (Fig. 7A). Cre-mediated recombination was verified by restriction fragment length polymorphism of the PCR products that were amplified from either the wild-type hcpA gene or the disrupted hcpA gene after Cre recombination (Fig. 7A). The resulting cell line, termed hcpA^lp, was then transformed with the hcpB targeting construct. Remarkably, we were not able to isolate any hcpB mutant clones in the hcpA^lp background from three independent transformations, indicating that disruption of the second hcp gene greatly affected viability. Ectopic overexpression of an His-HcpA protein in the hcpA^lp background (Fig. 7B) prior to hcpB targeting at least partially restored the disruption frequencies that we had observed in the Ax2 wild-type background (Table 2) and resulted in double-knockout strains. This suggested that the prior introduction of a functional gene could rescue the presumably lethal phenotype. The genetic composition of isolated clones was confirmed by PCR on the wild-type and mutated hcpB locus by PCR on the hcpA^lp locus and by Western blots showing expression of His-HcpA (Fig. 7C). No obvious phenotypic differences between the hcp^AB mutant-His-HcpA and the parental hcpA^lp-His-HcpA cells could be observed (data not shown). These findings indicated that the two Hcp isoforms were largely redundant and that the presence of at least one isoform was essential for cell viability. The data further confirmed that ectopically expressed His-HcpA can functionally replace the endogenous gene, at least with respect to cell viability.

Overexpression of GFP-HcpA, but not of GFP-HcpB, leads to an increased number of anaphase bridges.

The C-terminal GFP fusion HcpA-GFP was consistently expressed at much lower levels than HcpB-GFP, with almost no nucleoplasmic fluorescence (Fig. 2A). Although both constructs were driven by an actin 15 promoter, this difference could also be detected in cotransformants expressing HcpA-RFP and HcpB-GFP fusion proteins (Fig. 2B). In contrast, N-terminal GFP fusions of HcpA and HcpB showed approximately equal expression levels with similar nucleoplasmic background (Fig. 8A). Because of the almost identical expression levels, which were confirmed by Western blots (data not shown), we used the N-terminal GFP fusions for functional analysis.

Interestingly, when studying mitotic heterochromatin dynamics, we observed striking phenotypes that were particularly enriched in cell lines expressing GFP-HcpA.

Compared to the control transformation, we found a more than fourfold increase in DNA bridges between the separated DNA masses in late mitotic cells (Table 3).

Statistical evaluation of the data (see Materials and Methods) showed on the significance level of a maximum of 0.005 that overexpression of GFP-HcpA resulted in a higher frequency of anaphase bridges than in the other cell types.

Anaphase bridges were reminiscent of the abnormal separation of dicentric chromosomes in, for example, Drosophila melanogaster. GFP-HcpA fluorescence frequently localized to these DNA bridges (Fig. 8C), indicating that heterochromatin was involved in these structures. Furthermore, overexpression of GFP-HcpA caused pronounced growth defects in axenic suspension culture compared to control transformants (Table 1).

Overexpression of GFP-HcpB did not increase the number of anaphase bridges but still caused growth defects that were even more severe than those observed for GFP-HcpA-expressing strains (Tables 1 and 2). The results were reproduced with several independent clones, but the reason for slow growth could not be elucidated. In order to show that the recombinant proteins were indeed overexpressed compared to the wild type, we performed semiquantitative RT-PCR analysis of both strains. This showed that the mRNA levels of the respective Hcp isoforms were significantly increased in comparison to the endogenous transcripts (Fig. 8B). Although we could not test for expression levels of the endogenous protein, it is very likely that the recombinant Hcp isoforms were also more abundant than the respective wild-type proteins.

In order to gain further insight into the function of HP1 proteins in heterochromatin, we overexpressed truncated HcpA or HcpB protein lacking the C-terminal chromoshadow domain (HcpAΔC-GFP and HcpBΔC-GFP). We assumed that these proteins still localize to heterochromatin (73) but prevent the formation of higher order chromatin structures, since HcpAΔC-GFP was not able to dimerize in vitro (Fig. 3C). Hence, the mutant proteins should compete with both Hcp wild-type forms for binding sites and result in a dominant-negative loss-of-function phenotype when expressed in sufficient amounts.

The truncated proteins were, at least in part, localized to the same major heterochromatic focus as the full-length proteins. This was confirmed by colocalization with H3K9me2 staining and close proximity to the centrosome as identified by tubulin staining (data not shown). Staining of the minor subnuclear foci was, however, strongly reduced (Fig. 8A), suggesting distinct CSD-dependent and CSD-independent targeting of HP1 proteins in Dictyostelium. Overexpression of the truncated HcpA protein led to severe growth defects in axenic suspension culture (Table 1). Microscopic analysis of cells in suspension culture revealed that approximately half of the cells, compared to less than 5% of the control transformant, were abnormally rounded and had aberrantly condensed DNA. However, these did not represent prometaphase cells, since no significant mitotic defects could be detected (data not shown). Moreover, the plating efficiency on bacterial lawns of the HcpAΔC-GFP strain was more than threefold reduced compared to the control strain (Fig. 8D). Interestingly, overexpression of HcpAΔC-GFP, although expressed at levels similar to those of full-length GFP-HcpA, did not elicit increased numbers of anaphase bridges as seen for full-length HcpA (Table 2). It appears that overexpression of functional HcpA, which is able to interact with other proteins via its CSD, is required for this defect.

The decrease in cell viability of the HcpAΔC-GFP overexpression strain supported our assumption that the truncated protein had a dominant-negative function. This was, however, not as severe as a complete loss-of-function mutation in both Hcp isoforms.

RNAi-mediated heterochromatin formation in Dictyostelium?

In Drosophila melanogaster (62) and S. pombe (22), HP1 loss-of-function mutants influence H3K9 methylation levels and their subnuclear distribution. In Dictyostelium discoideum, the knockout mutants of individual HP1 genes did not cause significant changes in histone H3K9me2 subnuclear distribution (Fig. 9). This indicated that either the two isoforms are redundant for recruiting histone methyltransferases (HMTs) or that HMTs act upstream of HP1.

The RNAi machinery and their products, short interfering RNAs (siRNAs), are implicated in DNA methylation and in chromatin organization (22, 41). In Dictyostelium there is at least a correlation between the RNAi pathway and the silencing of a major heterochromatin component, the retrotransposon DIRS-1 (32). We therefore examined if RNA-directed RNA polymerases (RdRP), which are involved in RNA-mediated gene silencing, were required for heterochromatin formation in Dictyostelium. No obvious defects could be detected in H3K9 methylation (Fig. 9) or localization of the HP1 proteins (data not shown) in either of the three knockout strains of the RdRP genes rrpA, rrpB, and rrpC (38). Although we cannot rule out that in Dictyostelium RdRPs play a minor or a redundant role in heterochromatin formation, it is possible that, like in other organisms, there are parallel pathways for heterochromatin formation, including RNAi-independent ones (26).